https://www.cnas.org/publications/reports/future-proofing-frontier-ai-regulation
"…proliferation of any given level of #AIcapabilities will be delayed only a few additional years. At present, the cost of training models at any given level of #AI capabilities declines rapidly due to algorithmic progress alone…"
what we will have is a structural conflict of a #AIvAI: it's fucking already baked in.
where that might eventually lead is the real conundrum. pretzel logic.
at some point the AI-human interface may be far enough along that executive decisions *by humans* might be of value in real-time situations, where violent force occurs.
https://www.cfr.org/podcasts/killer-robots-and-autonomous-weapons-paul-scharre
whether that's good or bad is an even deeper question.
rexi
